A 99.2TOPS/W Transformer Learning Processor with Approximated Attention Score Gradient Computation and Ternary Vector-Based Speculation

Ping-Sheng WuYu-Cheng LinChia-Hsiang Yang2024-10-162024-10-162024-06-16https://scholars.lib.ntu.edu.tw/handle/123456789/722141A 99.2TOPS/W Transformer Learning Processor with Approximated Attention Score Gradient Computation and Ternary Vector-Based Speculationconference paper10.1109/vlsitechnologyandcir46783.2024.10631391